# Public readiness applicability rules — 1.2

Approved for subsequent assessments on September 21, 2026. This supplements decision-rules.md 1.1 without changing the 87 checks, point weights, cap rules, or existing published snapshots. Reports must state the rules version; do not silently relabel an older score.

## Determine the public task first

A missing required public capability is a finding on its applicable parent check. A feature must not become N/A merely because it is absent. Record the observed public purpose and the specific requirement.

A check about an object's syntax, internal links, parity, metadata or error behavior has no object to inspect when that optional object is absent. Mark that dependent check N/A with the absent object and its parent finding identified. Do not award half credit for a test that never occurred.

- No structured payload: schema syntax/type/ID/parity/current-fact/FAQ checks are N/A. Report the optional structured-data absence separately and unscored; this catalog has no separate schema-presence point. No new presence penalty or weight is introduced.
- No machine representation: grade applicable advertised-discovery coverage on `aeo.machine.alternate_advertised`. The representation's cleanliness, negotiation, parity, metadata and missing-resource children are N/A. If multiple representations exist, inspect them; never use this exclusion to hide a broken advertised link.
- No discovery map: grade applicability/absence on `aeo.discovery.map`. Map-link validity and boundedness are N/A without a map.
- No sitemap: preserve the sitemap/robots parent findings. Sitemap URL integrity, membership, lastmod and index-child checks are N/A without the corresponding object. Multi-page internal-link discovery remains relevant even without a sitemap.
- No canonical declaration: retain the page-level canonical finding. Fetching a declared canonical target is N/A if no target is declared. Other existing identity surfaces must still be compared where applicable.
- An orphan-comparison exclusion requires observed single-page public purpose and no separate public page inventory to compare. Missing sitemap data alone is insufficient for a multi-page product.

## Unknown and owner-only

Unknown means feasible public evidence is missing or unresolved. A publicly linked third-party destination is not owner-only merely because the auditor skipped it. Noncommitting public navigation may reveal fields and authentication requirements; stop before submissions, reservations, payments, or account changes.

Owner-only means the necessary evidence requires authentication, private deployment state, personal data, a submission/commit action, or an owner-controlled sandbox. Record that boundary. Observed labels or static constraints do not prove post-submit error handling.

## Event Timing resolution and delivery

The Event Timing API has a minimum 16 ms reporting threshold and 8 ms duration granularity. Observe both `event` (threshold 16) and `first-input`; the latter is observable below the event threshold. Retain trusted-input records, native interaction count where supported, visibility, browser conditions and all delivered entries. Flush the observer after painting and at the end; associate by event timestamps, not by callback arrival. A flush wait is never latency.

For the bounded lab path, group entries by nonzero interaction ID and take the maximum duration per interaction, then the maximum for each fresh run. Include first-input evidence. With complete observer support, three confirmed native/trusted interactions per run, successful state assertions and final delivery flushing, missing sub-threshold events remain censored below the reporting threshold; do not assign them zero. A confirmed absence of above-threshold entries is a reporting bound, not an exact duration. If instrumentation, eligibility, visibility, delivery or action completion is unproven, leave the result unknown.

Use the existing repeated-run lab decision bands; this is not field INP or a claim about real-user p75. The transfer/interaction composite still needs a separate justified resource assessment. Missing a documented byte budget does not by itself prevent a judgment: inspect actual blocking/resource-chain behavior and state the local decision rationale.

Sources: [W3C Event Timing](https://www.w3.org/TR/event-timing/) and [Google's INP guidance](https://web.dev/articles/inp). The dependency and scoring choices above are Lazyweb's audit policy, not a Google ranking rule.